In [1]:
# import needed libraries
import cv2
import matplotlib.pyplot as plt
import glob
1. Generating black and white sketch and Noise Reduction Using Median Filter¶
In [2]:
# default color is BGR
def plot_image(image,color,title):
plt.title(title)
if(color=="GRAY"):
plt.imshow(image,cmap='gray')
else:
plt.imshow(image)
plt.xticks([])
plt.yticks([])
In [3]:
image = cv2.imread('KATE.JPG')
In [4]:
# plot the original image as RGB
image_RBG = cv2.cvtColor(image,cv2.COLOR_BGR2RGB)
plot_image(image_RBG,"RBG","Original RGB Image")
In [5]:
image_gray = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
plot_image(image_gray,"GRAY","Original RGB Image")
In [6]:
# smoothing the image using median filter with window size 7
median_image = cv2.medianBlur(image_gray, ksize=7)
plot_image(median_image,"GRAY","Smoothed Grayscale Image")
2. Edge Detection Using Laplacian Filter¶
In [7]:
laplacian_image = cv2.Laplacian(median_image, ddepth=-1,ksize=5)
plot_image(laplacian_image,"GRAY","Edge Detection")
applying threshold with lower bound 125 and upper bound 255¶
In [8]:
ret,thresh = cv2.threshold(laplacian_image,125,255,cv2.THRESH_BINARY_INV)
plot_image(thresh,"GRAY","Image Thresholding")
3. Generating a color painting and a cartoon¶
In [9]:
# resize the image to half of its original size
resized_image = cv2.resize(image, (0,0), fx=0.5, fy=0.5)
plot_image(cv2.cvtColor(resized_image,cv2.COLOR_BGR2RGB),"RBG","Resized Image")
applying 7 iterations of bilateral fiter¶
In [10]:
diterations = 7
for i in range(iterations):
resized_image = cv2.bilateralFilter(resized_image, d=9, sigmaColor=9, sigmaSpace=7)
bilateral_image=cv2.resize(resized_image,(image.shape[1],image.shape[0]))
plot_image(cv2.cvtColor(bilateral_image,cv2.COLOR_BGR2RGB),"RGB","Bilateral Filter")
combine the edges withe the bilateral filtered image¶
In [11]:
cartoon = cv2.bitwise_and(bilateral_image,cv2.cvtColor(thresh,cv2.COLOR_GRAY2BGR))
plot_image(cv2.cvtColor(cartoon,cv2.COLOR_BGR2RGB),"RGB","Final Output")
4. Final Function¶
composite all steps in one function
- convert the image to gray
- reduce noise on the image using medianBlur filter
- use the function Laplacian() to detect edges
- put a threshold on the result and invert pixels
- apply 7 iterations of bilateral filtering
- combine the result cartoon effect with the original image using bitwise_and function and return the final result
In [12]:
def image_cartoonifier_converter(image):
image_gray = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
median_image = cv2.medianBlur(image_gray, ksize=7)
laplacian_image = cv2.Laplacian(median_image, ddepth=-1,ksize=5)
ret,thresh = cv2.threshold(laplacian_image,125,255,cv2.THRESH_BINARY_INV)
iterations = 7
resized_image = cv2.resize(image, (0,0), fx=0.5, fy=0.5)
for i in range(iterations):
resized_image = cv2.bilateralFilter(resized_image, d=9, sigmaColor=9, sigmaSpace=7)
bilateral_image = cv2.resize(resized_image,(image.shape[1],image.shape[0]))
cartoon = cv2.bitwise_and(bilateral_image,cv2.cvtColor(thresh,cv2.COLOR_GRAY2BGR))
return cartoon
In [25]:
def plot_side_by_side(img_1, img_2, size=(16,16)):
fig = plt.figure(figsize=(size))
fig.add_subplot(1, 2, 1)
plt.axis('off')
for i, img in enumerate([img_1, img_2]):
ax = fig.add_subplot(1,2,i+1)
img = cv2.cvtColor(img,cv2.COLOR_BGR2RGB)
plt.imshow(img)
plt.axis('off')
In [13]:
images = [cv2.imread(file) for file in glob.glob('test/*')]
In [15]:
i = 0
for image in images:
i += 1
cartoon = image_cartoonifier_converter(image)
cv2.imwrite("test/results/"+str(i)+".JPG", cartoon)
In [29]:
for image in images:
plot_side_by_side(image, image_cartoonifier_converter(image))
